Animal Learning & Behavior 1983, II (2), 223-228 .Two-choice, observational learning and reversal in the rat: S-S versus S-R effects

M. RAY DENNY, R. CHARLES BELL, and CARLA CLOS Michigan State University, East Lansing, Michigan

In Experiment 1, male rats were trained to press both bars in a two-choice apparatus and were then given observational training of a go/no-go discrimination in which the observed opera­ tion of two inaccessible, dissimilar bars by a hidden experimenter constituted S+ and 8-. After discrimination was established, individual rats were permitted access to the two bars. Six of the seven rats consistently pressed the S+ bar on 10 test trials, but failed to reverse bar pref­ erence after observational training was reversed. In Experiment 2, nine naive males received the same observational training as in Experiment 1, but without any pretraining to press either bar. All rats pressed the S+ bar on initial test and did so consistently throughout the 10 trials. Six of these rats received reversal training of the go/no-go discrimination after the 10 test trials. As in Experiment 1, all rats failed to press the new S+ bar. However, five of six rats in another group, which received reversal training prior to any test trials, did reverse and press the new S+ bar. In Experiment 3, controls for possible confounding effects of overtraining trials were conducted. These manipulations had no effect; the rats tested before reversal still failed to press the S+ bar, and the rats reversed before testing all reversed or pressed the most recent S+ bar. That is, S-R learning predominated over S-S learning if active, though unreinforced, responding to a particular bar intervened. In contrast, however, a cognitive (S-S) interpretation of directed response learning was supported by the results of Experiment 4, in which the rats that learned the go/no-go discrimination without responding (only by auditory and light cues) failed to press the S+ bar consistently.

Over the years it has become increasingly clear when permitted to do so fits a cognitive, or S-S, that the response-reinforcer contingency is not neces­ view of learning, somewhat reminiscent of Tolman's sary for learning an instrumental response (Lajoie (1932) "what leads to what" theory and largely in & Bindra, 1976). As these investigators point out, contrast to the results of the Spence and Lippitt "phenomena of response substitution (Lashley & (1946) latent learning studies. Here it should be em­ McCarthy, 1926), observational learning (Kohn phasized that the old latent learning studies tested & Dennis, 1972), and learning without responding for S-S relations that were irrelevant to the rat's (Dodwell & Bessant, 1960; Solomon & Turner, motivational state, whereas the present study is con­ 1962) demonstrate that a response can be acquired cerned with S-S learning that is relevant. without its reinforcement or even its occurrence In part, the present study can be conceptualized during the training trials" (p. 157). as contributing to an analysis of standard obser­ In the present study, the rat's observation of dis­ vational learning (Del Russo, 1975; Groesbeck & tinctive environmental events with definite hedonic Duerfeldt, 1971; Kohn & Dennis, 1972). Three cues consequences (stimulus sequences) does lead to the can be identified in the standard observational learn­ learning of the appropriate response in the absence ing procedure: the manipulandum, some movement of its occurrence. The rat first watches both positive or change in the manipulandum, and the model. and negative events occur, and demonstrates gol It may well be, however, that the model is a rel­ no-go discrimination that indicates it has learned atively unimportant factor. The observation of these S-S relationships. The learning of the appro­ changes in the manipulandum in correlation with priate response is then indexed by permitting the reinforcement or nonreinforcement may be sufficient animal to make either the response that should pro­ to elicit a full-blown directed response in subsequent duce the positive stimulus event or the one that test trials, as found by Jacobson and Sisemore (1976) should produce the negative event. That an obser­ in human subjects. vation-trained rat makes the appropriate response The standard procedure that evolved at the com­ pletion of Experiment 1 used completely naive rats, ones that had never pressed a bar (manipulandum) This research was supported, in part, by All University Research grants. Requests for reprints should be directed to: M. R. Denny, before. After these rats had learned a go/no-go dis­ Department of Psychology, Michigan State University, East crimination through observing that the experimenter­ Lansing, Michigan 48824. produced movement of one bar (S+) was associated

223 Copyright 1983 Psychonormc Society, Inc. 224 DENNY, BELL, AND CLOS with obtaining food and the movement of the other before they received observational training in the bar (8-) was not, they were given access to the bars tunnel. We were simply concerned with whether to see which one, if either, they pressed. Thus, the the rats would press the bar whose operation was present study can be classified as being similar to uniquely correlated with food. a discriminative autoshaping experiment in which responding to the discriminanda is blocked (Browne, Method Subjects. Seven hooded male rats, approximately ISO days 1976), or as a model-absent observational learning of age and from a colony maintained by the Department of study. Both paradigms share the same elements: Psychology at Michigan State University, were employed as the observation of contingencies and a transfer-of­ subjects. Hooded rats were selected because of their superior training test. vision. The rats were maintained at about 90070 of ad-lib weight and were 23 h deprived of food when run in the experiment. Apparatus. The two-choice observation enclosure is pictured EXPERIMENT 1 in Figure 1. The bars moved by the experimenter from behind the opaque end panel had certain stimulus consequences. The tip of the left bar had a total excursion of 8.5 em and made The general procedure becomes clear from a dia­ a sharp click when it struck the chamber floor. The activated gram of the apparatus used (Figure 1). The rat is right bar had an excursion of 2.2 em, turned on a dim light stationed in the hardware-cloth tunnel looking out above the bar, and produced a faint click from the microswitch. from behind the closed door at the two bars in the Preliminary training. Training the rats to press either bar to receive reinforcement at the end of the tunnel was a laborious test chamber. The hidden experimenter operates task. In nearly every case, the bars had to be baited with wet the bars individually from behind the opaque end mash and the wet mash gradually faded. Shaping was also em­ wall, and the operation of one particular bar (8+) ployed. Presumably the delayed reinforcement involved in en­ is consistently correlated with the presence of food tering the tunnel and traversing its length to the food tray as at the other end of the tunnel. After the rat has well as learning two responses more or less simultaneously made the learning difficult. All rats were trained until they pressed clearly learned to differentiate between the operation a bar within 10 sec of entering the test chamber, and they were of the two bars, it is permitted to enter the test cham­ equally often reinforced on each bar. To effect this equation, ber to see whether it will press the correct bar. a rat was occasionally prevented from having access to a par­ When we first began this research, we seriously ticular bar through use of a clear plastic barrier. Each rat was then given a test for bar preference with either 10 or 11 dis­ doubted that a naive rat that had learned to dis­ crete trials; in every case but one, a preference was determined criminate between the activations of the two bars within 10 trials, with a rat selecting one bar either 6 or 7 times. just by watching from the tunnel would actually In the observational training that followed, the S+ bar was al­ press either bar when they were made accessible ways the nonpreferred bar (four rats trained with S+ bar on during test trials. Consequently, the initial phase the right, three with S+ bar on left). All rats were quickly pre­ trained in the tunnel to leave the food tray and approach the of Experiment 1 involved the conventional training tunnel door by being given a 97-mg food pellet each time they of each rat to press both bars about equally well did this, for a total of five reinforcements. Obse"atlonal training. This phase began on the day follow­ ing the end of pretraining. Trials occurred only when the rat in the tunnel was looking toward the test chamber (Figure I). With very few exceptions, the rats promptly turned toward the tunnel door after visiting the food tray. On S+ trials of the go/no-go discrimination, a 97-mg Noyes food pellet was de­ livered only after the rat had jackknifed around in the tunnel en route to the food tray. This procedure guaranteed that the moving bars and attendant stimuli maintained exclusive stim­ ulus control, of the go/no-go behavior, since the slight noise of the pellet's striking the tray followed the response. Also, the intertrial interval (IT!) was a computer-generated Poisson dis­ tribution of time intervals, with a mean of 10 sec and a range of I to 60 sec, so wait-time could not become a reliable cue for leaving the tunnel door and approaching the food tray. The < ITI started after the rat returned to the door from the food tray or after the rat had waited 10 sec at the door on S- trials. Figure 1. The two-eholce obse"atlon apparatus. The fnnnel­ Extra visits to the food tray during the IT! increased the IT! only shaped compartment contalDlng ban Is caUed the test chamber. if an animal failed to return to the door by the end of the ITI, (A) The Uabt above right bar. (B) RIght bar. (C) Wooden block and post-ITI returns were never followed by an S+ or S- trial below right bar. (D) Left bar. (E) Hardware-elotb tunnel door until 3 to 5 sec had elapsed. In np position, when rat can leave or enter tunnel. Door Is down Trials were administered in blocks of 20, 10 S+ and 10 S­ when rat Is In tunnel watching ban being activated. (F) Hardware­ trials randomly distributed; generally, one or two blocks were cloth tunnel In which rat obse"es ban and Jackknifes around presented daily until the discrimination criterion was reached. to obtain food reward. (G) Food tray In which peDet is dropped The criterion was (1) inhibiting on 9 of 10 S- trials by con­ through tube. The celllng over the tunnel Is opaque; the one tinuing to face the test chamber for at least 10 sec after the over the ban Is transparent, letting in room ligbt. Tbe experi­ S- bar moved, and (2) leaving the tunnel door for the food menter, sitting almost opposite to the tip of the left bar, can see the tray in less than 10 sec on all 10 S+ trials of a block (well-trained subject at all times through the near wall opposite to the tunnel, subjects left in I or 2 sec). but the subject cannot see the experimenter untU after it turns As soon as the criterion was reached, the tunnel door was toward the food tray. (AU other walls are opaque.) raised, and when the rat entered the test chamber, the tunnel OBSERVATIONAL REVERSAL 225 door was closed behind it. As soon as a bar was pressed (turn­ trials through consistent responding in the test cham­ ing the light on for the right bar and deflecting the left bar to ber stimulus situation (Denny & Adelman, 1955). a horizontal position or below from a position .45 rads above horizontal), the rat was removed from the test chamber and re­ This learned response is protected from extinction turned to its home cage. Thus, with respect to the food tray, because reversal training in the tunnel is in a stim­ the rat was neither reinforced nor nonreinforced for barpressing ulus situation that is different from the situation during the test, although conditioned reinforcement could pre­ in which the barpressing occurred. For extinction sumably occur as a result of the light or sharp click. On each to take place, the original response and the com­ of 3 succeeding days, the rat was given 20 more observation trials followed immediately by a test, as described above. On peting response must occur in similar stimulus situ­ the next 3 days, the rat was tested directly before the 20 booster ations (Denny, 1971; Denny & Adelman, 1955). trials and directly after, yielding a total of 10test trials. The test of this interpretation was to reverse a group At the completion of the 10 test trials, all subjects that had on the go/no-go discrimination before it had re­ shown a clear tendency to press the S+ bar were given reversal ceived any test trials. Without test trials to mediate training in the tunnel; that is, the original S+ bar became the S- bar, and the original S- bar, the new S+ bar. Observational S-R learning, any response mediated. by the orig­ training continued until the same criterion as originally used inal S-S learning should be extinguished by learn­ was met, and then all subjects were administered 10 test trials, ing the S-S reversal. Thus, the hypothesis under test as described above. On early S+ trials, in reversal, in which the is that the new S+ bar is consistently pressed if subject inhibited for 10 sec, the S+ bar was reactivated and the rat always received reinforcement on S+ trials. all test trials follow the reversal training. To test for generality across strains, both hooded Results and Discussion and albino rats were used in Experiment 2. The seven rats learned the go/no-go discrim­ ination in a mean of 232 trials, including the criterial Method Subjects and Apparatas. Seven albino and five hooded male block. Six of them pressed the S+ bar on the ini­ rats, 165 days of age, were divided as equally as possible into tial test trial, and those six pressed the S+ bar on two groups of six rats each. Three additional males, two hooded an average of 83OJo of the 10 test trials. The other and one albino, were used to pilot the testing procedure. All rat (trained to right bar as S+) pressed the S- bar rats were under 23-h food deprivation and were maintained at 90"0 of ad-lib weight because more highly deprived rats bit the 80070 of the time. After reversal training in the tun­ tunnel door and visited the food tray excessively often. The nel, the six reversed rats persisted in pressing the apparatus was identical to that ofExperiment 1. old S+ bar on an average of 85OJo of the test trials, Preliminary trailliDI. Individual rats explored both halves even though the go/no-go discrimination reversal of the observation apparatus for approximately 10 min, with was learned in a mean of 210 trials. both bars inoperable. Ten 97-mg Noyes pellets were available in the food tray during exploration. Next, five additional food On the initial test trial, a rat typically explored pellets were individually delivered while the rat was confined the test chamber for a minute or two before press­ to the tunnel; each pellet was delivered only after the rat had ing a bar (the test chamber was funnel-shaped to left the food tray and faced the closed tunnel door before re­ minimize exploration), but by the last two test trials turning to the food tray (Figure 1). Observational traiDiq. This training, prior to test trials, was barpressing occurred within 3 to 25 sec and the identical to Experiment 1 for all 15 rats, except that half of the bar was typically depressed the full distance. Per­ animals in each group had the left bar as S+ and half had the formance on the 3 test trials given prior to the 20 right bar. For the six rats that were reversed before testing extra training trials was not degraded. (Group RBT), reversal training began on the day after the cri­ These generally positive results encouraged us terion was met and continued until this criterion was reached again. All rats received their first test trial on the criterial day; to eliminate the barpress pretraining in Experi­ the remaining nine trials were delivered as described for Ex­ ment 2 to see whether observational training alone periment 1. A second reversal followed the first set of test trials would produce appropriate responding during test and was carried to the same criterion. A second testing session in rats that had never pressed a bar. of 10 trials was then administered as before. The six rats tested before reversal (Group TBR) received all 10 test trials admin­ istered as in Experiment 1, that is, with 120 additional observa­ EXPERIMENT 2. tional learning, or booster, trials prior to reversal learning and the final 10 test trials. All test-trial sessions for both groups We first ran three naive rats with the new pro­ involved the 120 booster trials. Since the pilot animals that re­ ceived one or two test trials prior to reversal training performed cedure and discovered that they would press the close to chance on second testing, no test trials at all were given S+ bar as well as, if not slightly better than, the Group RBT before reversal training. rats of Experiment 1. By this time, an explanation of the reversed rats' failure to reverse bar prefer­ Results and Discussion ence during the second test session in Experi­ The original go/no-go discrimination was learned ment 1 became apparent. So these three rats were in a mean of 207.5 trials (N =15), the first reversal used to pilot the critical manipulation. We then in a mean of 209.6 (N =15), and the second in a designed a two-group experiment to test our hy­ mean of 196.6 trials (N = 5). All nine rats tested pothesis. The hypothesis involved the following before reversal pressed the S+ bar on the initial points. Rats learn a particular barpress during test test trial and pressed the S+ bar on an average of 91OJo 226 DENNY, BELL, AND CLOS of all the test trials. Group TBR pressed the S+ reached the reversal criterion, ETC received its initial set of bar on an average of 87070 of the test trials during the 10 test trials and then was reversed and tested again. This final first test session and the same bar on an average of step indexed the effect of the initial testing period. There were two tested before reversal (TBR) groups, one without the extra 82070 of the test trials after reversal training. That training, or booster, trials during the initial testing phase (TBR­ is, as in Experiment 1, all subjects of Group TBR no extra) and one with the 120 extra training trials as in Experi­ failed to reverse bar preference. Five of the six rats ment 2 (TBR). All trials were administered in the same fashion in Group RBT pressed the new S+ bar on an average with the same ITI as in Experiment 2. Reversal training and sub­ sequent testing were specifically the same as in Experiment 2, that of 92% of the 10 test trials, while one rat (hooded) is, with booster trials given during test days for all groups in order never pressed it. After the second reversal, these to maximize the possibility of reversing bar preference. same five rats failed to reverse, pressing the new S- bar on an average of 96070 of the 10 test trials. Again, intervening barpressing during test trials, Results and Discussion which presumably involved S-R learning, since la­ All three groups showed similar observational tencies declined steadily over test trials, prevented learning, with means to criterion of 196.7, 196.7, reversal of bar preference. With Fisher's exact test, and 213.3 on the original go/no-go discrimination the difference between groups on the 10 test trials and means of 250, 216.7, and 220 on first reversal following the first reversal was highly significant for ETC, TBR-no extra, and TBR, respectively. (p = .00(75). In the only group that received a second reversal If the data from both Experiments 1 and 2 are (ETC), the customary trend toward facilitation combined for the critical initial test trial following was apparent (mean = 156.7 trials). the original go/no-go discrimination, 15 of 16 rats All six rats in Group ETC pressed the new S+ ~ar pressed the S+ bar (p = .00026 by the exact binomial on the initial test trial, and on the average pressed test). It 93.3070 of the 10 test trials, clearly showing re­ In short, albinos showed model-absent observa­ versal of the directed response. Thus, the 120 over­ tionallearning roughly to the same degree as hooded learning trials did not prevent reversal. When ETC rats, and the hypothesis that S-S reversal learning was reversed for the second time on the go/no-go of bar preference occurs only when intervening discrimination, this time after 10 test trials, it failed barpressing on test is eliminated was supported. to reverse bar preferences, as was true of all groups However, since Group TBR had 120 more trials tested before reversal, pressing the new S+ bar on on the original go/no-go discrimination than an average of only 10% of the 10 test trials. Group RBT before it received reversal training, Five of six TBR subjects pressed the S+ bar on there is the possibility that overtraining could have the first test trial. However, for the first time in produced the failure of Group TBR to reverse bar this type of experiment, two male rats in the TBR preference. This would seem to be very unlikely group responded at chance level to the bars on ini­ because overtraining, if anything, typically has the tial testing, yielding as a group 66.7070 pressing of opposite effect, facilitating, rather than hindering, the S+ bar overall. One of these two animals re­ reversal learning. But a direct test of this possibility ~ersed bar preferences after reversal training, press­ was in order. mg the same bar as the last three barpresses on the initial set of 10 test trials at a 100010 level. The rat EXPERIMENT 3 that did not reverse also pressed, at a 100010 level on the second test, the same bar as the last three To control for the 120 extra trials' possibly con­ barpresses of initial testing (the last three barpresses founding effects on reversal performance, a com­ happened to be consistently to one particular bar plete experiment was run. To control this factor for both subjects). In other words, it appears that in the two ways possible, a group that was reversed three unreinforced barpresses in a row can yield before testing also received an additional 120 trials sufficient S-R learning in the test chamber via con­ on the original go/no-go discrimination, and a tiguit~ or elicitation principles (Denny, 1971; group that was tested before reversal received no Guthrie, 1952) to mediate consistent responding extra trials during the initial testing phase. thereafter (Denny, 1959). The other four rats in TBR continued to press Method the original S+ on an average of 97.5010 of the 10 Subjects, Apparatus, aDd PreUmiaary TniDiDg. Eighteen al­ test trials after reversal training, failing like all other bino males, approximately 110 days of age, served as subjects. TBR groups to reverse bar preference. Likewise, The apparatus and preliminary training were as in Experiment 2. all six rats in the TBR-no extra group pressed the ObservaUoDal traiDiDI. In Experiment 3, there were three S+ bar on the initial test trial and on 86.7010 of the groups of six rats ~~h. ~ extra trials control (ETC) group received 120 overtraining tnals, 20 per day, after reaching orig­ first set of 10 test trials, and then pressed the old inal criterion. These trials carne before the rats were reversed S+ on an average of 95010 of the 10 test trials fol­ and before any testing. After reversal training and after they lowing reversal training. In short, TBR-no extra OBSERVATIONAL REVERSAL 227

and TBR, groups that differed only on whether or to respond and when an S- barpress was counted not they received the 120 extra trials during testing, as an error, exactly 50070 of the test trials of the did not differ significantly on reversal of bar pref­ controls were incorrect. Thus, it appears that moving erence (Mann-Whitney U test, p> .53), while TBR bars were critical to observational learning of press­ and ETC, which differed only on whether or not ing the S+ bar. This finding makes it difficult to they received test trials prior to reversal training, explain barpress learning noncognitively-on the did differ significantly on reversal performance basis of incipient approach to S+ while watching (Mann-Whitney U test, p < .02). Overtraining, or from the tunnel and accidentally pressing the S+ extra trials, is irrelevant for reversal of bar pref­ bar during test trials. erence. GENERAL DISCUSSION EXPERIMENT 4 Probably the most interesting aspect of these re­ In addition to controls for the 120 extra trials, sults is that a small amount of S-R learning appears there was another important control that needed to predominate over a considerable amount of S-S to be conducted. There was the distinct possibility learning, at least as we interpret it. According to that a rat in the test chamber simply approached the elicitation framework (Denny, 1971; Denny & the region where the S+ bar was located and, in Adelman, 1955), reversal failure after testing is not so doing, accidentally pressed this bar. It is possible unexpected because during observational reversal that the rat does not see the bars move during ob­ training in the tunnel the rat is not learning any servational training and is not attempting to make response that directly competes with a particular the S+ bar move during test. Observation of the barpress response in the unique stimulus situation rats' behavior during the later stages of observa­ of the test chamber. In an autoshaping context, when tional training indicated, however, that the rats were rats are permitted to make a competing response in attending closely to bar movement; they would often the same situation in which learning occurred (Davey, jackknife around as soon as the S+ bar started to Oakley, & Cleland, 1981), they do, in fact, reverse. move and well before any other visual or auditory The similarity to autoshaping and sign tracking cue occurred. But a more direct test was in order. (Hearst & Jenkins, 1974) has already been noted, and it stands repeating. But there seem to be cer­ Method tain differences: (1) The learned response appears Subjects, Apparatus, and PreHminary Training. Six hooded full blown, occurring on the initial test trial; it just males, about 140 days of age, constituted the bar-movement does not facilitate the learning of the operant as in control group. The apparatus and preliminary training were as in Experiment 2. autoshaping of blocked responses (Browne, 1976). Observational training. Three rats learned the go/no-go dis­ (2) The nature of the learned response seems to be crimination with the light above the right bar going on and the specified by the nature of the S+ (moving bar), faint sound of a microswitch click as the S+ (no bar moved) although considerably more research is needed to and with the sound of a sharp blow (click) on the base of the back panel directly behind the point where the tip of the left establish this point. (3) The learned response, once bar would ordinarily strike the floor as S- (no bar moved). it is made, is never associated with reinforcement, For the other three rats, S+ and S- were reversed, and no bar yet it persists (Experiment 3). (4) There is a critical moved. Testing was the same as in Experiment 2, and if a sub­ two-choice test phase. (5) When the rat presses the ject did not press either bar within 5 min after entry to the test S+ bar, it does attempt to get back in the tun­ chamber it was removed. This procedure had never been neces­ not sary in previous experiments. Because our interest lay only in nel to obtain a food pellet; the rat presses the bar determining whether the moving bars were critical to the original rather "deliberately" and just stays there, often observational learning of pressing the S+ bar, the subjects in hanging onto the bar. This last point invites the Experiment 4 were not given reversal training. speculation that the rat is acting as if it were com­ menting on the environment, telling us in its own Results and Discussion way that this is the S+ bar. The bar movement controls learned the go/no-go discrimination in a mean of 212.5 trials, about the same number as the subjects trained with moving REFERENCES bars. But their test trial behavior was markedly dif­ BROWNE, M. P. The role of primary reinforcement and overt ferent. For the first time on test trials, including movements in autoshaping in the pigeon. Animal Learning & the initial one, the rats often failed to respond within Behavior, 1976,4,287-292. 5 min after entry to the test chamber, and the la­ DAVEY, G. C., OAKLEY, D., & CLELAND, G. G. Autoshaping in tencies when barpressing occurred were also long. the rat: Effects of omission on the form of the response. Journal The median latency was 136 sec; for six hooded ojtheExperimentalAnalysisojBehavior, 1981, 36, 75-91. DEL Russo, J. E. Observational learning of discriminative rats trained with moving bars in Experiment 2, the avoidance in hooded rats. Animal Learning & Behavior, 1975, median was 49 sec. Also, when there was a failure 3,76-80. 228 DENNY, BELL, AND CLOS

DENNY, M. R. One bar-press per day: Acquisition and extinction. KOHN, B., & DENNIS, M. Observation and discrimination learn­ Journal of the Experimental Analysis of Behavior, 19S9, 1, ing in the rat: Specific and nonspecific effects. Journal of 81-8S. Comparative and Physiological Psychology, 1972, 71, 292-296. DENNY, M. R. Relaxation theory and experiments. In F. R. Brush LAJOIE, J., & BINDRA, D. An interpretation of autoshaping and (Ed.), A versive conditioning and learning. New York: Academic related phenomena in terms of stimulus incentive contingencies Press, 1971. . alone. Canadian Journal ofPsychology, 1976, 30, 151-173. DENNY, M. R., & ADELMAN, H. M. Elicitation theory: I. An LASHLEY, K. S., & McCARTHY, D. A. The survival of the maze analysis of two typical learning situations. Psychological Re­ habit after cerebellar injuries. Journal of Comparative Psy­ view, 19",61,290-296. chology,1926,6,423-433. DODWELL, P. C., & BESSANT,D. E. Learning without swimming SoWMON, R. L., & TURNER, L. H. Discriminative classical con­ in a water maze. Journal of Comparative and Physiological ditioning in dogs paralyzed by curare can later control dis­ Psychology, 1960,53, 422-42S. criminative avoidance responses in the normal state. Psycho­ GROESBECK, R. W., & DUERFELDT, P. H. Some relevant vari­ logicalReview, 1962,69, 102-219. ables in observational learning of the rat. Psychonomic Science, SPENCE, K. W., & LIPPITT, R. An experimental test of the sign­ 1971,11,41-43. Gestalt theory of trial and error learning. Journal of Experi­ GUTHRIE,E. R. Thepsychology oflearning (rev. ed.), New York: mental Psychology, 1946,36,491-502. Harper & Row, 1952. TOLMAN, E. C. Purposive behavior of animals and men. New HEARST, E., & JENKINS, H. M. Sign-tracking: The stimulus­ York: Appleton-Century, 1932. reinforcer relation and directed action. Austin, Tex: Psycho­ nomic Society, 1974. JACOBSON, J. M., & SISEMORE, D. Observational learning of a lever pressing task. Southern Journal ofEducational Research, (Manuscript received May 21, 1982; 1976, 10, '9-73. revision accepted for publication February 22, 1983.)